NEW YORK, NY – During a record-breaking year for holiday travel and with airfare prices outpacing inflation, consumers find it difficult to find a travel itinerary that matches their needs. We are all familiar with the process of being overwhelmed by the number of possible flights and rapidly changing prices when searching for a flight. However, shifting data privacy rules and ever-changing consumer needs make it harder for travel platforms and other online businesses to tailor their offerings to customers and make the search process easier for consumers. A new Columbia Business School study develops and tests a brand-new approach that helps companies predict customers’ needs in real time as they progress in their purchase journey by analyzing customers’ most recent searches, clicks, and purchases on their website with other customers’ behaviors. This approach is suitable for situations where the company does not have any data on customers prior to visiting the website. It is ten times more accurate in understanding customer needs early on in the journey (e.g., first query) than existing methods, boosting click-through rates on the first page of search by as much as 28%. Furthermore, the proposed approach has meaningful privacy implications as it does not require saving customers’ data beyond the current journey.
For the study, The Customer Journey as a Source of Information, Professor Oded Netzer, the Arthur J. Samberg Professor of Business, and his co-authors, London Business School Professor Nicolas Padilla and Harvard Business School Professor Eva Ascarza, test their machine learning model on a major online travel platform using data from 4,500 customers who searched for flight tickets between May and November 2017. They analyze customer “journeys,” or customers’ interactions with the business, including what they search for and click on when searching for a flight. Their model links three types of information together to predict customers’ preferences: shoppers’ behavior within their current journey, their behavior during previous journeys, and similar patterns from other shoppers, continuously updating as shoppers search for products. Their application of the model illustrates that companies can show customers products that better match their needs by positioning products in a way that better resonates with the customers’ journey and also improves engagement by allowing companies to infer preferences like airline alliances, number of stops, and price sensitivity. As the customer continues to engage with the platform the model learns what the customer is looking for. After five clicks on the platform the model accuracy increased to 73% versus 25% after the first two clicks. Thus, the model helps customers find what they want quicker, without relying on customer stored historical data, which poses privacy concerns.
As shoppers are increasingly price-sensitive due to increasing costs and regulators restrict third-party data available to companies and marketers, businesses should find a way to “have the customer at hello” and show the customer what they need quickly and effectively. This new study offers a new approach that allows companies to do exactly that in a way that respects their privacy.
To learn more about the cutting-edge research being conducted, please visit Columbia Business School.
###
